VizChitra 2025

This project, led by a team at Dvara Research, aims to support high-level decision-making by leveraging a Market Monitoring Tool that uses social media data to surface customer grievances related to the Unified Payments Interface (UPI). Given UPI’s vast reach—handling over 10 billion transactions in a single month—even an industry-leading error rate can impact tens of millions. Traditional grievance channels often fail to capture the true scale of issues due to drop-offs across the redressal pipeline. Recognizing this gap, the project taps into publicly available complaints on platforms like X (formerly Twitter) and the Google Play Store.

The tool uses Natural Language Processing (NLP) techniques to categorize complaints, detect overlapping issues, measure provider responsiveness, and track how user hurdles evolve over time. These insights help uncover systemic problems—such as repeated transaction failures or poor grievance redress—and assess how different UPI providers handle customer dissatisfaction.

Key Takeaways:

1. Social media can serve as a powerful, real-time data source for identifying systemic issues in financial services, especially in high-volume platforms like UPI.
2. A complaint is rarely just a complaint—it often reflects the hurdles faced by millions who don’t or can’t voice themselves publicly.
3. The combination of machine learning and natural language tools offers a scalable, nuanced way to detect patterns that traditional redress mechanisms, particularly in legacy systems, might miss.
4. But it’s not enough to process the data—the stories embedded in these complaints matter just as much.

The audience includes developers, data scientists, designers, and fintech practitioners who are building user-focused technology. We hope to present this tool to the community to gather feedback on improving our tech stack, refining our NLP pipeline, and designing more insightful and actionable visualizations.

Speaker: Dwijaraj heads the Financial Systems Design Initiative at Dvara Research, where he works on addressing systemic challenges in the financial sector, often through data and technology. He has worked with governments, regulators, and self-regulatory bodies on a range of issues, including supervisory tech, overindebtedness, and regulatory design. His recent work includes building machine learning models for text categorization and debt distress prediction. He blends public policy insight with hands-on technical expertise—and has probably spent more time cleaning datasets than making coffee.